Assuring Quality of Experience for IPTV

IP television has emerged as a key value proposition and differentiator for telecommunication service providers in the ever-growing competition with cable and satellite operators for delivery of triple-play services. IPTV and video on demand promise a personalized user experience by letting users decide what they watch, and when, through a user-friendly electronic programming guide. However, cable and satellite TV have set the bar high on service reliability and availability, and potential users expect IPTV to meet or exceed this quality of experience (QoE), all at a comparable price.

While traditional best-effort IP services such as Web browsing and e-mail can adapt to network congestion, video, voice and interactive multimedia services can become seriously degraded. Critical success factors to address this include:

A reliable and QoS-enabled network foundation. Unless the network is highly reliable with built-in QoS mechanisms, achieving a reliable QoE is unattainable.

Network design and dimensioning. It’s important to optimize and rightsize the network to handle anticipated service demand and minimize congestion risks.

Integrated service admission control. This ensures network capacity usage stays within the limits the network was designed for, even during extreme peak conditions.

End-to-end service assurance. This refers to providing an effective means to measure service demand and network performance and verify that QoE objectives are met.

Click to Enlarge

Figure 1 depicts a triple-play service delivery architecture based on IP over Ethernet to provide a reliable and QoS-enabled network foundation. Key attributes to optimize performance and reliability are:

• Distributed policy enforcement in combination with centralized policy management optimizes control over the end-to-end service delivery path and gives the flexibility and elasticity to adapt to changing traffic patterns and service mix over time.
• Hierarchical VPLS and hierarchical QoS give service separation and fine-grain resource control per subscriber and per service with bandwidth preserving scheduling.
• Redundancy implemented at the node, card, link and connectivity levels protects against single point of failures and supports nonstop service delivery and upgrades.

Some examples to minimize video delivery cost are:

• IP multicast replication for broadband TV in access, aggregation and edge nodes instead of only at a centralized node maximizes multicast transport efficiency.
• IGMP snooping in the access and aggregation network prevents the replication and forwarding of unwatched TV channels, saving cost while maximizing channel bouquet.
• Subscriber- and service-aware aggregation allow insertion of regional broadcast channels, advertisements and popular VoD content closer to end users, instead of backhauling all video content over a centralized service edge and IP/MPLS core network.

It is essential to properly dimension network capacity to ensure a low congestion probability for voice and video services during normal operational conditions.

Click to Enlarge

Figure 2 depicts the potential resource contention points in the video delivery path, from left to right:

• The first mile — the access loop — typically is configured to enable concurrent usage of all services. DSL loop analyzers can make sure that the training rate is conditioned at stable levels for which capacity commitments can be made for subscriber services.
• The second- and third-mile are the most important contention points to watch, as here the aggregate amount of video traffic easily may dominate network traffic as indicated by the two bandwidth cylinders at the right. Second-mile bandwidth consumption can fluctuate dynamically while it is relatively costly and difficult to add more capacity.
• The video servers are dimensioned for a maximum amount of streams that can be supported concurrently, and the links to connect these servers to the network can be statically provisioned to support the maximum amount of stream bandwidth.

Service Admission Control (SAC) determines whether sufficient network resources are available to admit a new service request. If service demand is higher than the available network capacity can handle, SAC will deny new service requests to prevent congestion and degradation of existing services. Receiving a “network busy tone” is not a good user experience and a lost revenue opportunity, so proper network dimensioning must ensure this does not occur frequently. Although SAC clearly is compromising between service quality and availability, it may be too costly or nearly impossible to ensure the network always will have enough capacity to handle every possible demand peak or failure scenario.

SAC applies for services with deterministic QoS and throughput needs, because shaping and policing of individual packets to handle congestion seriously would degrade quality. Applying SAC for video services is very challenging because video traffic is very bursty, with large rate variations between near-still images and fast-action sequences. In the case of IPTV-based broadcast video, it also must be ensured that the introduction of SAC will not add significant latency to the channel-change mechanism. For VoD, it must be taken into account that titles may be rented for several days and sessions can be paused, resumed, rewound or forwarded at any time.

Click to Enlarge

Figure 3 shows the service assurance and control architecture to address these issues:

• Application-level admission control between set-top boxes (STB) and IPTV middleware (dotted horizontal green arrow) limits the total amount of standard and high-definition video streams a home may receive concurrently (first-mile constraint).
• Network-level service admission control for IPTV is enforced at the broadband service aggregator node (BSAN), broadband service aggregator (BSA) and broadband service router (BSR) and triggered by channel change requests (i.e., IGMP joins) to limit the aggregate link bandwidth (second- and third-mile constraints) of broadcast TV (BTV) channels while minimizing additional channel change latency.
• VoD session admission control is enforced by a centralized AAA/SAC policy decision point that monitors aggregate traffic load on the second- and third- mile links and interfaces with the IPTV middleware to limit aggregate VoD session bandwidth.
• Service assurance is a centralized application that collects and correlates diagnostic information from network elements, STB and video servers (yellow labels).

The effectiveness of SAC depends on the level of control intelligence and integration achieved:

• Subscriber location and entitlements, SLAs, network topology and resource state are important inputs to make intelligent, policy- based admission control decisions that maximize revenue from available network capacity.
• Integrating with all elements in the service delivery network from subscriber equipment to video middleware allows a service provider to execute policy decisions effectively while providing a meaningful “busy tone” or alternative to end users when a service request is denied.

A key strength of IPTV is the wealth of monitoring and control information that can be collected, aggregated and correlated to determine video and audio quality on a network-wide and per-user basis. By taking QoE measurements from all elements in the service delivery path, it becomes possible to correlate transport-layer QoS and application QoE to determine in how far bit errors and packet loss that occurred during network transport can be compensated at the application layer by means of error correction algorithms or packet retransmission.

Measuring historical usage patterns and trends enables providers to accurately dimension network capacity and prevent issues.

Arnold Jansen is the product and solution marketing manager for Alcatel. He can be reached at Arnold.Jansen@alcatel.com.

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines